Report on the 2nd ACM SIGIR/SIGKDD Africa School on Machine Learning for Data Mining and Search
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
EVENT REPORT
Report on the 2nd ACM SIGIR/SIGKDD
Africa School on Machine Learning for Data
Mining and Search
Tanya Berger-Wolf Ben Carterette Tamer Elsayed
The Ohio State University Spotify Qatar University
berger-wolf.1@osu.edu carteret@acm.org telsayed@qu.edu.qa
Maria Keet Fabrizio Sebastiani
University of Cape Town ISTI-CNR
mkeet@cs.uct.ac.za fabrizio.sebastiani@isti.cnr.it
Hussein Suleman
University of Cape Town
hussein@cs.uct.ac.za
Abstract
We report on the organization and activities of the 2nd ACM SIGIR/SIGKDD Africa School
on Machine Learning for Data Mining and Search, which took place at the University of
Cape Town in South Africa January 27–31, 2020.
1 Introduction
In January of 2019, ACM SIGIR and ACM SIGKDD co-sponsored the 1st SIGIR/SIGKDD Africa
School on Machine Learning for Data Mining and Search1 , which took place at the University of
Cape Town in South Africa. The event was conceived with the aims of increasing opportunities in
research from traditionally underserved communities, growing the IR and data mining communities
in sub-Saharan Africa, and expanding the horizons of IR and data mining research. The school
took place for the second time in January of 20202 , again at the University of Cape Town. This
report is on the organization and activities of AFIRM 2020. For more background and context
for the event, please refer to the report on SIGIR 2019 by Carterette et al. [2019].
AFIRM 2020 grew in many ways from AFIRM 2019. It went from single-track to parallel-
track; both the number of students and the number of instructors increased; and topics were more
diverse.
1
http://www.sigir.org/afirm2019
2
http://www.sigir.org/afirm2020
ACM SIGIR Forum 1 Vol. 54 No. 1 June 20202 Event Organization
2.1 Organizing team
The organizing team consisted of the following committees and individuals:
1. Steering committee
• Ben Carterette, Chair (Spotify)
• Diane Kelly, SIGIR Representative (University of Tennessee Knoxville)
• Jian Pei, SIGKDD Representative (Simon Fraser University, British Columbia)
• Abdiganyi Diriye (IBM Kenya)
• Ricardo Baeza-Yates (NTENT / Northeastern University)
• Lighton Phiri (University of Zambia)
• Reuben Dlamini (University of the Witwatersrand)
• Douglas Oard (University of Maryland)
2. General chairs
• Tamer Elsayed, co-Chair (Qatar University)
• Hussein Suleman, co-Chair (University of Cape Town)
3. Local support at the University of Cape Town
• Mary-Jane Antia
• Jecton Anyango
• Catherine Chavula
• Sarah Dsane-Nsor
• Frances Gillis-Webber
• Zola Mahlaza
• Jivashi Nagar
• Toky Hajatiana Raboanary
• Joseph Telemala
• Tezira Wanyana
4. Program committee
• Ricardo Baeza-Yates (NTENT / Northeastern University)
• Maria Maistro (University of Copenhagen, Denmark)
• Douglas Oard (University of Maryland)
• Iadh Ounis (University of Glasgow, UK)
ACM SIGIR Forum 2 Vol. 54 No. 1 June 2020• Lighton Phiri (University of Zambia)
• Barbara Poblete (University of Chile)
• Matteo Riondato (Amherst College, USA)
• Mark Sanderson (RMIT University, Australia)
• Guido Zuccon (The University of Queensland, Australia)
2.2 Event funding
ACM SIGIR and ACM SIGKDD funded the event with USD20,000 each. Additional support for
local arrangements came from the University of Cape Town. We also received sponsorship from
the South Africa National Research Foundation (Gold), Web4Good (Gold), Microsoft (Silver),
Google (Bronze).
2.3 Participation
We issued open calls for course proposals and for general attendance. The CfP for course proposals
was published 7 July 2019, and proposals were accepted through 2 August 2019. There were several
rounds of calls for attendance, the first for those that wished to apply for travel support, followed
by additional calls for unsupported participation.
2.3.1 Courses and instructor participation
We received 31 course proposals, more than twice as many as 2019. These were reviewed by the
program committee listed above. Final decisions were made after discussion among the six authors
of this report over the course of two conference calls. The decision criteria included having a good
mix of data mining and search topics, having some introductory courses as well as more advanced
courses, ability to program courses in parallel, and fit with the rest of the program.
The final list of courses accepted through proposal and invitation were:
1. Introduction to Information Retrieval, Charlie Clarke and Zeynep Akkalyoncu Yilmaz
2. Data Mining: From an ad-hoc art form to solid, reproducible processes leading to transparent
and accountable outcomes, Andreas Rauber
3. The Role of Evaluation for Search, Nicola Ferro and Maria Maistro
4. Fact Checking, Axel-Cyrille Ngonga Ngomo
5. Crowdsourcing for Language and Multimedia Research, Gareth Jones and Maria Eskevich
6. Spatio-Temporal Data Mining: Methods and Applications to Ocean Data, Stan Matwin and
Luis Torgo
7. Learning to Rank for Information Retrieval with Neural Networks, Bhaskar Mitra, Nick
Craswell, Daniel Campos, and Emine Yilmaz
ACM SIGIR Forum 3 Vol. 54 No. 1 June 20208. Explainable Machine Learning for Trustworthy Artificial Intelligence, Fosca Giannotto and
Dino Pedreschi
There were also three invited keynote lectures:
1. IR4All: Building Search Engines for Everyone, by Doulgas W. Oard
2. IR and DM for Development, by Hussein Suleman
3. NLP for African Languages, by Maria Keet
2.3.2 Student participation
Calls for participation were issued in two phases after the program had been finalized. The first
call was for applications for travel support. We received 273 applications.
We allocated USD25,000 for supporting participant travel, including accommodation at the All
Africa House just off the University of Cape Town campus. As we wanted to ensure this would be
a truly multi-national event, we tried to allocate awards such that those receiving support would
receive nearly all of what they needed for their travel—it is not common for African institutions to
support international travel, so many of them would not be able to attend at all unless their travel
was fully supported. We also wanted a diverse event, in terms of nationality, gender, background,
and other considerations.
We offered support to 24 applicants. Nearly all accepted the support by registering for the
event. Most were able to attend, but five who accepted support were ultimately unable to attend
for various reasons. Nigerian attendees who were offered support the previous year but not able
to attend were given priority for support this year, and most of them were able to attend.
A second call, without travel support, was issued closer to the dates of the event. Several
dozen additional attendees, mainly from South Africa, registered via this call.
In the end there were 50 participants representing 10 different African countries: Botswana,
Cameroon, Ethiopia, Ghana, Malawi, Morocco, Namibia, Nigeria, South Africa, and Zambia. This
is an 85% increase in participation over AFIRM 2019. Additionally, gender representation was
greatly improved over AFIRM 2019, with 20 women included in the 50 participants.
3 The event
The majority of the event took place in lecture rooms and labs in the Computer Science Building
at the University of Cape Town. There were two dedicated lecture halls and two dedicated labs.
Each day was split between lectures and labs, with lectures starting at 0900 and going through
1200, at which point lunch was served. Lab sessions picked up at 1300 and went through to 1630.
After a break, each day concluded with a reception, keynote, or panel.
The first two days of the event were single-track, featuring lectures and labs to introduce
participants to the search and data mining settings for the rest of the week. Each day after that
had two courses in parallel, one for a data mining track and one for a search track.
ACM SIGIR Forum 4 Vol. 54 No. 1 June 2020Day 1 The event kicked off with the University of Waterloo team Charlie Clarke and Zeynep
Akkalyoncu Yilmaz given an introduction to Information Retrieval via lectures and hands-
on labs. The day concluded with a keynote lecture from Doug Oard on building search
engines for under-resourced languages.
Day 2 Next on the agenda was an introduction to data mining by Andreas Rauber and Tanya
Berger-Wolf. An evening poster reception allowed instructors and students to meet, discuss
the students’ work, and get to know each other.
Day 3 The third day of the event was the first with parallel tracks. Nicola Ferro and Maria
Maistro presented lectures and hands-on labs on topics in search evaluation, while Axel
Ngongo gave lectures and labs on fact checking. Wednesday officially closed with a keynote
lecture from Hussein Suleman on IR and DM for development. An optional panel on ap-
proaches to research took place later at All Africa House, where many participants were
staying, moderated by Lighton Phiri and Ben Carterette.
Day 4 The fourth day continued the split, with lectures and labs on crowdsourcing given by
Gareth Jones and Maria Eskevich in one room, and lectures and labs on spatio-temporal
data mining in another. The final keynote lecture by Maria Keet closed the day’s sessions,
and this was followed by the gala dinner at the Gold Restaurant in Cape Town.
Day 5 The week ended with lectures and labs on learning to rank given by Bhaskar Mitra,
Nick Craswell, Emine Yilmaz and Daniel Campos, and explainable machine learning by
Fosca Giannotti and Dino Pedreschi. The school concluded with a panel featuring four
of the schools attendees—Ahmed Olanrewaju of the University of Ibadan, Nigeria; Chi-
tundu Chisunka-Mwila of the University of Zambia; Sylvester Hatsu from Accra Technical
University, Ghana; and Meryem Janati Idrissi of Mohammed VI Polytechnic University,
Morocco—and moderated by Tamer Elsayed, and a closing ceremony.
4 Conclusions
The 2nd ACM SIGIR/SIGKDD Africa School on Machine Learning for Data Mining and Search
ran from January 27–31, 2020, in Cape Town, South Africa. We increased diversity of both
presents and participants in terms of subject matter, origin, and gender, all of which we had as
goals. Response was overwhelmingly positive for the second time, and we are highly motivated to
continue the initiative.
4.1 Next steps
We continue to maintain the open-subscription Google Group at afirm-africa@googlegroups.
com for communication among those who have attended AFIRM 2019 or 2020 in person as well as
any one else who is interested in joining. Unfortunately, it is likely that AFIRM will not happen
in January 2021 due to travel restrictions and bans on gatherings as a result of the COVID-19
pandemic.
ACM SIGIR Forum 5 Vol. 54 No. 1 June 2020References Ben Carterette, Suleman Hussein, and Oard Douglas W. Report on the 1st acm sigir/sigkdd africa school on machine learning for data mining and search. SIGIR Forum, 53(1):3–13, 2019. ACM SIGIR Forum 6 Vol. 54 No. 1 June 2020
You can also read
